Cuda.is_available() in MXNet

Do we have

torch.cuda.is_available()

equivalent in MXNet?

There is a workaround

from mxnet import npx
npx.num_gpus()

But wanted to know if there is a straight-forward API exposed for the same; else I can open feature request on Github repo

Just wondering, what information do you expect from a is_cuda_available function that you don’t get from num_gpus()?

That there are NVidia GPUs available?

regards,
Lieven

Same info as num_gpus but just that it is consistent with other framework APIs and is rather direct way
using num_gpus seemed a bit round about.

try:

mx.test_utils.list_gpus()
1 Like